The Value of Technically Poor Recordings for Tweaking

[fas42]

 

As promised elsewhere, a thread to explain my thinking ...

 

@MLXXX's post, and my responses,

 

Quote

 

I understand that may be a long-held view of yours fas42, so you are unlikely to resile from it now.

 

However, are you able to explain it in a way that is easier to understand, because applying the view literally to a very poor quality source would tend to hide rather than reveal the effect of a tweak, as discussed below.  Surely you are not suggesting use of very poor quality sources.

 

I am indeed, if by poor quality sources you mean the recording waveform, as against a poor quality source component.

 

Quote

 

*   *   *

 

 

Consider the following three, admittedly extreme,  circumstances:

 

1. Evaluating the tweak for reduction in harmonic distortion but the source is riddled with harmonic distortion (e.g. a 78rpm shellac disc from 1915).

 

Fundamentally, you don't worry about what sort of distortion you are trying to reduce - any gains, in any area, is a win. In simple terms, you want 100% of the waveform's embedded distortion to be audible, with 0% added by the playback chain. Why does this work? Because, the intermodulation between the recording's embedded distortion, and the playback's distortion, is the killer problem ... that is what has to be eliminated !!!

 

Quote

2. Evaluating the tweak for reduction in noise but the source is very noisy (e.g. a compact audio cassette from 1970).

 

Again, as above. Only the cassette noise, in its full glory, should be audible.

 

Quote

3. Evaluating the tweak for its ability to smooth the upper frequencies but the source is a recording of a standard AM tuner where high frequencies becomes progressively more severely attenuated starting at around 3kHz.

 

No 'smoothing' required - if you want the playback to be true what the AM sounded like, leave it alone. If you want a more correct FR, apply some DSP either in real time, or remaster the track and save offline, as the one to be played.

 

Quote

 

Taking each of the above in turn:

harmonic distortion in the source would swamp any additional harmonic distortion introduced by the audio system with or without tweak

 

noise in the source would mask noise introduced by the audio system with or without tweak

the lack of high frequency content in the source would preclude evaluating the high frequency performance of the audio system with or without tweak.

 

Not how it works - see above.

 

Also consider that the waveform in the source, no matter how distorted, always carries information about the captured event. Distortion by the playback carries zero information about the event; it treats every recording evenhandedly - in the quite typical scenario, a rig imparts a distinctive, repeating signature to every track; an ideal playback chain creates a sound world unique to the particular track being played.

 

Swing orchestra tracks from the 1930's are a good test: the brass section choruses, coming from the back of the soundstage, will sound "rough as guts" if a rig is poorly optimised. As the tweaking improves matters, this will steadily move towards sounding like an ensemble of musicians doing their thing; what you hear will become a meaningful part of the whole.

 

 

[POV]

POV has left the chat

[fas42]

What one has to deal with in audio, is the "uncanny valley" of perception. Well understood in the visual world, https://en.wikipedia.org/wiki/Uncanny_valley, it's almost unheard of in the sound game.

 

Where so systems end up, is somewhere in that dreaded trough, the valley - they struggle to sound good on anything but the "very best" recordings - what one has to understand is that there is a slope leading out of that 'bad' place - but the first step is to be aware that this is very much how it is.

 

'Poor' recordings expose how zombie, creepy the rig has become - it takes a concerted effort to realise the next stage of system refinement, to climb out of the valley ...

 

 

[Steff]

43 minutes ago, fas42 said:

Well understood in the visual world

 

Not at all "well understood" and even under contestation as an actual phenomenon (cf. the very wikiP article you link to). From that you extrapolate a whole auditory set of effects that you seek to map onto recording quality. In evidence of your initial hypothesis.

It's a little bit like stating that the Flying Spagetti Monster is proof of the Flat Earth theory.

[fas42]

56 minutes ago, Steff said:

 

Not at all "well understood" and even under contestation as an actual phenomenon (cf. the very wikiP article you link to). From that you extrapolate a whole auditory set of effects that you seek to map onto recording quality. In evidence of your initial hypothesis.

It's a little bit like stating that the Flying Spagetti Monster is proof of the Flat Earth theory.

 

Well, if you going to come from the angle that the uncanny valley in the visual world doesn't exist - try feeding that line to the movie makers who have had their very expensive CGI productions fail at the box office, because their humanoids fell into that bad place - then there is not much I can do for you, .

 

It took me quite some time to work out, for myself, what was going on - a recording which could send you from the room in disgust, when heard on some ambitious rig, became a complete enthralling musical experience when replayed with sufficient accuracy - how could this be?!! The answer is mighty simple - the human mind can compensate for losses of quality in what the senses pick up - but like most things, there's a limit. Fall under what's needed, and the mind says, chuck it!! If it didn't work this way, we would probably go mad trying to deal with all the real world sensory input, in our daily lives ...

[MLXXX]

21 hours ago, fas42 said:

As promised elsewhere, a thread to explain my thinking ...

 

@MLXXX's post, and my responses,

 

First off, I must thank you fas42 for following through on your promise by setting up this thread. Also I see that you helpfully split up my post into chunks and commented on each of those.

 

I will keep my comments in this thread relatively brief as I understand this subject area has been gone over in the past by others, in other forums, where you have been involved.  As best I can recall I haven't been involved myself in the past.

 

So some comments:-

 

 

21 hours ago, fas42 said:

I am indeed, if by poor quality sources you mean the recording waveform, as against a poor quality source component.

 

Yes, basically,  I was referring to the content of the recording itself being inherently technically poor by modern standards.

 

For example, a modern CD compilation of recordings of the famous Italian tenor Enrico Caruso would involve poor frequency response, high harmonic distortion and high background noise as the recording technology at the beginning of the 20th century leading up to Caruso's death in 1921 was very poor by today's standards.   (In practice, a CD compilation of old 78rpm discs will try to alleviate background noise and ameliorate the frequency response, but the poor technical quality will remain unmistakable.)

 

 

21 hours ago, fas42 said:

Fundamentally, you don't worry about what sort of distortion you are trying to reduce - any gains, in any area, is a win. In simple terms, you want 100% of the waveform's embedded distortion to be audible, with 0% added by the playback chain. Why does this work? Because, the intermodulation between the recording's embedded distortion, and the playback's distortion, is the killer problem ... that is what has to be eliminated !!!

 

The thing is that playback distortion these days is primarily in the speakers. The distortion in a modern system prior to the speakers is very small and so human hearing won't notice it.* 

 

Theoretically IMD contributed by speakers could arise from distortion in the source content, but the distortion in the source would need to be pretty high for that to be audible. It could be in early to mid-20th century recordings released on CD, or if playing back certain discs on a turntable particularly 78rpm shellac discs, or even modern 33 1/3 rpm vinyl discs with heavily modulated inner grooves.

 

_____________

 

* A major exception to that is the use of vinyl technology. There is considerable potential for vinyl playback to create additional distortion depending on the condition of the stylus, the playing weight, the degree of record wear, etc..  

 

21 hours ago, fas42 said:

Again, as above. Only the cassette noise, in its full glory, should be audible.

 

If you whisper in a quiet room at night, you may be heard by a person 10 metres away (if their hearing is intact).  However , if you whisper in a discotheque when the music is playing, you will not be heard by a person even one metre away.  So it would be when adding whisper quiet system noise to playback of a compact cassette.  It is inaudible.

 

Actually today system noise is typically even quieter than a whisper. 

 

Modern psychoacoustic codecs work on the basis of dropping low level sounds that the algorithms can detect but which human hearing cannot. At high bitrates such codecs become transparent to the human ear.

 

If you have not already tried out your ability to tell whether you can hear a difference between an original file and a high bitrate lossy version of it, fas42, you should. It is remarkable how poor human hearing is at detecting low level sound in the presence of high level sound.

 

 

21 hours ago, fas42 said:

No 'smoothing' required - if you want the playback to be true what the AM sounded like, leave it alone. If you want a more correct FR, apply some DSP either in real time, or remaster the track and save offline, as the one to be played.

 

I was referring to AM radio as an example of a source with no content above a certain audible frequency. AM radio as broadcast in Australia in the medium wave band is by law not permitted to have content above 9kHz, so as to avoid interference to other radio stations. In practice, high frequencies are severely attenuated well below 9kHz in consumer AM radios so as to reduce distracting noise (static etc).  The attenuation in the radio receiver begins at quite a low frequency, around 3kHz,  Wideband AM radios are extremely rare.

 

Obviously (to my mind), one would not use an ordinary AM radio to evaluate the high frequency performance of a speaker system.  One would stand a much better chance of evaluating high frequency loudspeaker performance using an FM radio as the source.

 

 

I think I'll leave my comments at that.   All the best to you, @fas42!

Edited August 19 by MLXXX

[fas42]

1 hour ago, MLXXX said:

 

For example, a modern CD compilation of recordings of the famous Italian tenor Enrico Caruso would involve poor frequency response, high harmonic distortion and high background noise as the recording technology at the beginning of the 20th century leading up to Caruso's death in 1921 was very poor by today's standards.   (In practice, a CD compilation of old 78rpm discs will try to alleviate background noise and ameliorate the frequency response, but the poor technical quality will remain unmistakable,)

 

My reference here is a CD of Nellie Melba tracks - starting in 1912. Sub-par reproduction is a caricature of an opera singer, a quaint, historical momento; but at its best, you "hear" Nellie, and note exactly how far back the piano was, accompanying her. More recent tracks easily point out how the technology improved in her lifetime; but, you are still able to easily hear her mastery of singing, right in the earliest ones - you don't shrink from the sound - you embrace it ...

 

1 hour ago, MLXXX said:

 

 

 

The thing is that playback distortion these days is primarily in the speakers. The distortion in a modern system prior to the speakers is very small and so human hearing won't notice it.* 

 

Completely opposite thinking, here. A poor electronics chain makes it trivially obvious that you are listening to equipment; and at the other end, an illusion is thrown up, which is impossible to break - one tiny weakness is enough to go from one standard of listening to the other.

 

1 hour ago, MLXXX said:

 

If you whisper in a quiet room at night, you may be heard by a person 10 metres away (if their hearing is intact).  However , i you whisper in a discotheque when the music is playing, you will not be heard by a person even one metre away.  So it would be when adding whisper quiet system noise to playback of a compact cassette.  It is inaudible.

 

Actually today system noise is typically even quieter than a whisper. 

 

The system noise I talk of is that which degrades the cassette tape noise. Consider listening through the finest headphones to specifically the noise of a cassette tape, and then a recording of that noise which is played back through your system, captured and played over the same headphones - you should not be able to tell which is the direct, and which has come via a system playback.

 

1 hour ago, MLXXX said:

 

If you have not already tried out your ability to tell whether you can hear a difference between an original file and a high bitrate lossy version of it, fas42, you should. It is remarkable how poor human hearing is at detecting low level sound in the presence of high level sound.

 

Been there, done that ... . I was curious whether MP3, using the LAME encoder, could "fool me" ... nope! No matter what settings I used, at the highest possible rates, it was always obvious that they differed, and each combo of settings of the encoder changed the character of the "tells". The MP3 didn't sound worse, just different - there was no winner ...

 

Of interest, this was done by burning WAVs of the original and MP3 variations, on a CDR; and playing that on a Sharp boom box - no difficulty in hearing the signatures.

 

1 hour ago, MLXXX said:

 

I think I'll leave my comments at that.   All the best to you, @fas42!

 

Thank you for your good thoughts .... cheers, Frank 

Edited August 19 by fas42

[MLXXX]

1 hour ago, fas42 said:

Been there, done that ... . I was curious whether MP3, using the LAME encoder, could "fool me" ... nope! No matter what settings I used, at the highest possible rates, it was always obvious that they differed, and each combo of settings of the encoder changed the character of the "tells". The MP3 didn't sound worse, just different - there was no winner ...

 

Of interest, this was done by burning WAVs of the original and MP3 variations, on a CDR; and playing that on a Sharp boom box - no difficulty in hearing the signatures.

It could be a while ago that you did that. An efficient codec today at 320kbps is very hard to pick. 

[fas42]

16 hours ago, MLXXX said:

It could be a while ago that you did that. An efficient codec today at 320kbps is very hard to pick. 

 

Encoders have settled down, for quite a while - the LAME stuff was all done at the 320bps level, full bandwidth, with lots of fine tuning, max quality settings tried. Opus encoding is probably the pick of the bunch, but still might be audible - a Round Tuit thing ...

[MLXXX]

1 hour ago, fas42 said:

 

Encoders have settled down, for quite a while - the LAME stuff was all done at the 320bps level, full bandwidth, with lots of fine tuning, max quality settings tried. Opus encoding is probably the pick of the bunch, but still might be audible - a Round Tuit thing ...

 

A way to test oneself rigorously and efficiently is with automated ABX software such as foobar 2000 (after installing its ABX plug-in).   If what you claim could be confirmed, it would appear your hearing is atypical, corresponding to say <1% of the population of adults with healthy hearing.*

 

Anyway, the degree of transparency of codecs is of course not your main point. Your main point as per the thread title is to recommend use of relatively poor quality material to assist in evaluating a hi-fi system, or as you put it in another thread:

 

"It's quite obvious a lot of people have little understanding of how to evaluate tweaks  - you never use your best recordings ... you always use your 'worst' ones - the latter will make it very clear whether anything in the sound has changed - and then you have to evaluate whether it was for the better, or for worse."

 

_________

* Up at 320kbps with an efficient codec It is not normal for codec artefacts to be in the category of "it was always obvious that they differed".  More typically the audible difference, if perceived, is quite subtle, and apparent only with particular selections of music and at particular times. Also it may only emerge for a listener if they have the advantage of immediate A  B comparisons.  Even then the difference from use of the the codec does not necessarily create the impression of a downgrade in quality as distinct from just being slightly "different".   YouTube uses around 130kbps.

 

 

Edited August 20 by MLXXX

[fas42]

39 minutes ago, MLXXX said:

 

A way to test oneself rigorously and efficiently is with automated ABX software such as foobar 2000 (after installing its ABX plug-in).   If what you claim could be confirmed, it would appear your hearing is atypical, corresponding to say <1% of the population of adults with healthy hearing.*

 

Anyway, the degree of transparency of codecs is of course not your main point. Your main point as per the thread title is to recommend use of relatively poor quality material to assist in evaluating a hi-fi system, or as you put it in another thread:

 

"It's quite obvious a lot of people have little understanding of how to evaluate tweaks  - you never use your best recordings ... you always use your 'worst' ones - the latter will make it very clear whether anything in the sound has changed - and then you have to evaluate whether it was for the better, or for worse."

 

_________

* Up at 320kbps with an efficient codec It is not normal for codec artefacts to be in the category of "it was always obvious that they differed".  More typically the audible difference, if perceived, is quite subtle, and apparent only with particular selections of music and at particular times. Also it may only emerge for a listener if they have the advantage of immediate A  B comparisons.  Even then the difference from use of the the codec does not necessarily create the impression of a downgrade in quality as distinct from just being slightly "different".   YouTube uses around 130kbps.

 

 

 

Interesting experience with foobar2000 and its ABX plug-in, years ago. Things may have changed, but the actual player, on the particular machine, produced very low quality SQ. Tried several other players, found a winner - all were better than foobar.

 

And then, discovered the methodology of the ABX beast was completely useless - it made temporary copies, with resampling, of what was supposed to be compared; A and B were now, Y and Z - ensuring results were meaningless ...

[MLXXX]

7 minutes ago, fas42 said:

And then, discovered the methodology of the ABX beast was completely useless - it made temporary copies, with resampling, of what was supposed to be compared; A and B were now, Y and Z - ensuring results were meaningless ...

I don't think foobar interferes unless replay gain settings are used. 

 

The main issue is to ensure the two files under comparison are perfectly time-aligned. 

 

Foobar2000 has been used for years on HydrogenAudio forum, as an approved form of evidence of successful ABX testing,  and they are very picky there on technical matters.

[fas42]

3 hours ago, MLXXX said:

Up at 320kbps with an efficient codec It is not normal for codec artefacts to be in the category of "it was always obvious that they differed".  More typically the audible difference, if perceived, is quite subtle, and apparent only with particular selections of music and at particular times. Also it may only emerge for a listener if they have the advantage of immediate A  B comparisons.  Even then the difference from use of the the codec does not necessarily create the impression of a downgrade in quality as distinct from just being slightly "different".   YouTube uses around 130kbps.

 

Yes. When I want to use YouTube material I access the Opus encoding, which has a full 20kHz bandwidth, at 160kbps - this is adequate for most listening, etc.

 

Depends upon what one wants to find out ... to see if there are obvious differences, between some tracks, I just time align them as channels in Audacity, and flick between them with the Solo button, while in Play. One can use the Repeat function to narrow in on some area, and play a sound pattern endlessly, to pick what may be there, that matters.

 

The first time I used the ABX in Foobar2000, I noted that replay in that mode sounded quite different from just normal playback ... what was going on?!! Digging around, I discovered what was happening "behind the curtain" - for casual use, it may have some value; but not something to be taken seriously ...

Edited August 20 by fas42

[MLXXX]

28 minutes ago, fas42 said:

Depends upon what one wants to find out ... to see if there are obvious differences, between some tracks, I just time align them as channels in Audacity, and flick between them with the Solo button, while in Play. One can use the Repeat function to narrow in on some area, and play a sound pattern endlessly, to pick what may be there, that matters.

That is quite a useful method.